Anthropic leaks 512,000 lines of Claude Code source
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic confirmed it accidentally leaked part of Claude Code's source code via an NPM package, calling the incident 'a release packaging issue caused by human error, not a security breach'
- PC Gamer reports approximately 512,000 lines of Claude Code's CLI source code were exposed, a figure echoed by TechRadar and rounded to 500,000 by Axios
- Anthropic stated that 'no sensitive customer data or credentials' were exposed in the leak, per the company's official response
- The New Stack reports the leaked source reveals 'swarms, daemons, and 44 features Anthropic kept behind flags' — exposing previously unreleased product capabilities
- The Register frames the incident as revealing 'how much info Anthropic can hoover up about you and your system' — a privacy dimension that goes beyond the code exposure itself
- Decrypt notes the leaked code is now permanently public — 'The Internet Is Keeping It Forever' — and cannot be recalled
Why it matters: Anthropic's AI coding assistant codebase — including 44 unreleased features and specifics of its data collection capabilities — is now permanently public. Competitors gain free access to architectural details of one of the most-used AI coding tools, while the Register's framing surfaces a privacy dimension that Anthropic's 'human error' message sidesteps.
